On Tue, Sep 6, 2022, at 11:52 PM, Daniel Vetter wrote:

On Sat, Aug 06, 2022 at 07:32:24PM +0300, Markuss Broks wrote:

Ok I have a more fundamental issue with this than the lack of proper patch
splitting I mentioned in the other thread.

This is the wrong place.

drivers/video/console is about the various vt console implementations,
which supply a struct consw to con_register_driver.

You are right, that was my mistake. The original patch was for
drivers/video/fbdev/, and I suggested moving it out of there because
it does not depend on the fbdev subsystem, but clearly my suggestions
of drivers/video/console was equally wrong.

This otoh is an (early) kernel/printk console implemented using struct
console. Totally different thing, and really shouldn't end up in
drivers/video/console imo. Somewhere in drivers/firmware might still be
the best place, the sysfb stuff is also there. Maybe
drivers/firmware/sysfb_earlycon.c?

drivers/firmware/ is better but doesn't sound great to me either,
since one important thing the patch does is to not make it depend
on EFI firmware or BIOS style screen_info any more. 

Maybe drivers/tty/earlycon_simplefb.c would work better, keeping
it close to the earlycon base support in drivers/tty/serial/,
the vt console and the old simplefb driver/, without tying to too
closely to fbdev of UEFI.

     Arnd